Questions lead to study designs

What is the problem?
What information is missing from existing knowledge?
What is my contribution to new knowledge

Epidemiological Study designs

Case Series
Cross sectional Surveys
Case Control Study
Longitudinal Studies
Cohort Studies
Randomised Controlled Trials

Case Series (Surveillance)

Surveillance
A number of individual cases are thoroughly described
No comparison group

Surveillance for Asthma

Results of Surveillance for Asthma

Results of Asthma

When is Case Series a Good Strategy?

When you want to find out incidence/prevalence of a condition
When you want to find out the prognosis of a disease
In studying an outbreak
In generating hypotheses

Time Series Studies

Some changes over time is modelled
Example: air pollution and deaths in Brazil
Number of daily deaths
Indicators of pollution on that day
Limitation: Cannot be extrapolated to individuals

Example of a Time Series Study

Brazil Air Pollution Study

Methods of the Air Pollution Study

Methods of a Time Series Study

Results of the Air Pollution Study

Results of Air Pollution Study

Graph of the Association between Air Pollution and Mortality

Results of Air Pollution Graph

Ecological Fallacy

Findings of an ecological study cannot be extended to individuals
Correlation does not imply causation

When is Time Series a Good Idea?

Initial exploration of a theory
Individual data on variables are impossible to obtain (when??)
When data are easily available over the Internet on aggregate measurements

Cross sectional Surveys

What is

Features of a Cross Sectional Survey

Snapshot in Time
One time process of obtaining data
Exposure and outcome data are taken at the same time
Good for studying prevalence of a condition
A cause and effect relationship cannot be deduced

Example of a Cross-Sectional Survey

Roro Study

Exposure in Roro to Asbestos

Exposure to Asbestos

Presentation of Results in a Cross sectional Survey

Presentation of Results Table

When is a cross-sectional survey a good idea?

Individual based study
Limited budget and time (relatively inexpensive study design)
Exploration of the health effect
When you are interested in studying prevalence of a condition
Large national and multinational surveys are useful

When is Cross-sectional survey not a good idea?

When you are interested in causal research
If you want to study prognosis of a health State
If you want to study incidence rather than prevalence of a condition

Case Control Studies

Conceptual Diagram o Case Control Study

Features of Case control studies

Cases and controls are compared with each other
Odds of exposure are measured
Odds Ratio is the measure
Logistic Regression is the statistical tool to be used
These studies are always retrospective
Good for rare disease/conditions/health states

Concept of Case Control Study

Condition	Cases	Controls
Exposed	A	B
Non-exposed	C	D

Odds of Exposure among cases = A / C Odds of Exposure among controls = B / D Odds Ratio (OR) is A * D / B * C (Cross Product)

Can you calculate the Odds Ratio?

Condition	Cases	Controls
Exposed	60	20
Non-exposed	40	80
Total	100	100

Calculation of Odds Ratio

Odds of Exposure Among Cases = 60/40

Odds of Exposure among Controls = 20/80

Odds Ratio = 60 * 80 / 40 * 20 = ??

What is your interpretation?

Example of a Case Control Study

Arsenic Methylation and Bladder Cancer study

Result of the Bladder Cancer Case Control Study

Odds ratio of bladder cancer for methylation

How did the risks change with Smoking and Gender

Confounding by Gender and Smoking

When is Case Control Study a Good Idea?

Limited budget and short Time
If the outcome is a rare disease/condition
If you are interested to study many exposures at the same Time
Individual data based causal research
You can "nest" case controlled studies within larger studies

Cohort Studies

Conceptual Diagram

Features of Cohort Studies

Exposure and non-exposure cohorts
Follow up over time
Always prospective
Relative Risk or Rate Ratio measured
Survival Analysis/Cox PH Model/ANOVA used for Analysis

Dunedin Longitudinal Study in New Zealand

Dunedin Study

Example of An Interesting Cohort Study from Dunedin Cohort

TV viewing

Results of the TV viewing and Health Study

TV view results

When is Cohort Study a great idea?

Strong study design for causal research
Good for rare exposures and common outcomes
When you are interested in multiple outcomes
When you want to measure Relative Risk or Rate Ratio
When you are interested to find out incidence of disease
When you want to study natural history of a health condition or exposure

Intervention Research: Randomised Controlled Trials

RCT conceptual

Features of Randomised Controlled Trials

Intervention and Control Groups
Random Allocation
Blinding
Follow up in time
Rates of outcomes or harms observed in both Groups

When is an RCT good idea?

For all intervention research (individuals and groups)
Closest to Experiment
Cox PH Model
Analysis of Variance

Problems with RCTs

Strong design for causality
Not generalisable
Small sample size
Not always relevant for health Studies
Expensive

What type of Study to conduct

What is the research question?
How much resource?
What is the outcome?
What results are needed?
What is the status of the knowledge?

Evidence Value Graph

Summary

A very high level overview of the research methods
Case series through RCTs
Many more interventional and observational study designs possible
Your choice will depend on the resources at hand and your research questions
A cross sectional survey is a very common study design in Health Sciences
Case control studies are also quite popular in Epidemiology
Large cohorts such as Dunedin Longitudinal studies enable cohort studies

Research Cycle